Re: [translate-pootle] Still problems committing with git

2008-07-03 Thread Christian Perrier
Quoting Lars Kruse ([EMAIL PROTECTED]):

  I think these are very valid comments. This whole conversation is about
  a bug anyway, and git is (as you mention) previously unsupported. If the
  code to support older git versions is available, we obviously use it,
  but I'm not too concerned about halting anything just to support all
  versions.
 
 I agree. Thus I will add a module for git1.5.2 within the next days.
 I hope, that it will work for git 1.4.x, too.
 If it does not work for git 1.4.x, then we will just ignore this issue until
 people report real demand, I guess.


All this seems fair. As Miklos pointed and when it comes at Debian,
using a recent Pootle version needs to install either a backported
package or a package from Debian lenny (which is what we do on
i18n.debian.net). So, having to install a backported git is not that a
big deal.

In short, if supporting git versions prior to 1.5 was possible, it
would be good. If that's complicated and involve losing valuable
developer time, it's not worth it.

Actually, people building Pootle production servers with Debian should
indeed use Etch plus Pootle from testing and git from www.backports.org




-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread Christian Perrier
Quoting Lars Kruse ([EMAIL PROTECTED]):

 C)
 - applying Miklos patches to the current git support
 - ignoring incompatibility with git=1.5.2
 
 Personally I would prefer (A) over (B) due to the cleaner code (in the
 long-term). But of course, I would not mind (B), too.
 The current version of git in debian stable seems to imply that (C) is not a
 good option for us, or?

We worked this around by using a more recent version of git, for
sure. But, if there's any way to support older versions, it would
certainly be better as you will at some moment definitely find someone
who prefers sticking with supported version of packages and software
on their distro.

So, if it is possible to support git versions below 1.5.2, it would
get my own vote.



-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread Christian Perrier
Quoting F Wolff ([EMAIL PROTECTED]):

 From Christians' feedback, I understand that things are well tested, and
 that we are happy with this working. (I haven't tested this yet.)

s/well tested/tested

:-)

We don't have anything in prodcution yet (but I now really work on
this, for translations related to the Debian Installer--you can for
instance begin some work on Afrikaans translation of D-I in Pootle if
you would like to help...:-))

So,  as all this is not used in prodcution, I can't tell this is well
tested.




-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread Miklos Vajna
On Wed, Jul 02, 2008 at 02:36:34AM +0200, Lars Kruse [EMAIL PROTECTED] wrote:
 1) the current implementation does not work without the --work-tree 
 parameter
 (e.g. with the two patches that were sent by Miklos (30.6.2008))
 
 2) git v1.5.3 (or higher) behaves as expected - it is specified/documented to
 need both --git-dir and --work-tree for any operation outside of the
 current directory
 
 3) for git v1.5.2 (and before) we have to change to the (base of the?) local 
 git
 repository before executing git

Right. the work-tree/git-dir switches allow out-of-tree git commands
which I did not need when I initially did Git support, and unfortunately
I forgot about them.

 C)
 - applying Miklos patches to the current git support
 - ignoring incompatibility with git=1.5.2

I think the situation where compabitility with git=1.5.2 makes sense is
Debian stable. It has Pootle 0.10.1. It just does not have Git support.
If a sysadmin decides to install newer a never (it was called
unofficial) version of Pootle, then I think it is not that hard to
install a newer version of Git as well.

I think if we want to support Debian stable, then we have to support
1.4.4.4, since that's the official version Debian comes with. There are
major changes between 1.4.x and 1.5.x, and - to be honest - I'm not too
motivated to write code for 1.5.x.

So I would vote for C).

Just my two cents.


pgpmyXzRn776n.pgp
Description: PGP signature
-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread F Wolff
On Wo, 2008-07-02 at 23:18 +0200, Miklos Vajna wrote:
 On Wed, Jul 02, 2008 at 02:36:34AM +0200, Lars Kruse [EMAIL PROTECTED] 
 wrote:

...

  C)
  - applying Miklos patches to the current git support
  - ignoring incompatibility with git=1.5.2
 
 I think the situation where compabitility with git=1.5.2 makes sense is
 Debian stable. It has Pootle 0.10.1. It just does not have Git support.
 If a sysadmin decides to install newer a never (it was called
 unofficial) version of Pootle, then I think it is not that hard to
 install a newer version of Git as well.
 
 I think if we want to support Debian stable, then we have to support
 1.4.4.4, since that's the official version Debian comes with. There are
 major changes between 1.4.x and 1.5.x, and - to be honest - I'm not too
 motivated to write code for 1.5.x.
 
 So I would vote for C).
 
 Just my two cents.

I think these are very valid comments. This whole conversation is about
a bug anyway, and git is (as you mention) previously unsupported. If the
code to support older git versions is available, we obviously use it,
but I'm not too concerned about halting anything just to support all
versions.

Friedel


-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread Lars Kruse
Hi,

I just committed Miklos' patches for the git module as r7710.
This should fix problems with git v1.5.3 (or higher).
(I did it, since I did not notice any objections ...)


  I think if we want to support Debian stable, then we have to support
  1.4.4.4, since that's the official version Debian comes with. There are
  major changes between 1.4.x and 1.5.x, and - to be honest - I'm not too
  motivated to write code for 1.5.x.
  
  So I would vote for C).
  
  Just my two cents.
 
 I think these are very valid comments. This whole conversation is about
 a bug anyway, and git is (as you mention) previously unsupported. If the
 code to support older git versions is available, we obviously use it,
 but I'm not too concerned about halting anything just to support all
 versions.

I agree. Thus I will add a module for git1.5.2 within the next days.
I hope, that it will work for git 1.4.x, too.
If it does not work for git 1.4.x, then we will just ignore this issue until
people report real demand, I guess.

Lars

-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-02 Thread F Wolff
On Do, 2008-07-03 at 01:56 +0200, Lars Kruse wrote:
 Hi,
 
 I just committed Miklos' patches for the git module as r7710.
 This should fix problems with git v1.5.3 (or higher).
 (I did it, since I did not notice any objections ...)

Perfect! Thank you everybody for helping out to solve and test this.

Friedel




-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-01 Thread Miklos Vajna
On Mon, Jun 30, 2008 at 11:35:13PM +0200, Miklos Vajna [EMAIL PROTECTED] 
wrote:
 here is a patch, on top of my previous one.

Friedel, if you could review/commit these 2 patches, I would appreciate
it. I think they are definitely useful not only for Debian but for
anybody who use po files outside the Pootle directory (we at Frugalware
do not use external po files, that's why I haven't hit this bug yet).

Thanks.


pgpsx7iiSoMmP.pgp
Description: PGP signature
-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-01 Thread Miklos Vajna
On Tue, Jul 01, 2008 at 09:03:37PM +0200, Wynand Winterbach [EMAIL PROTECTED] 
wrote:
 Miklos, I'll also review your changes tomorrow and try to check them in as 
 soon as I can. If Friedel is slow to respond, it's because he's at a 
 conference this week :).

Oh, OK, I was not about to hurry you, just wanted to point out these
patches were not meant to specific to the Debian guys. :-)


pgpztyATXdC8u.pgp
Description: PGP signature
-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-01 Thread Lars Kruse
Hi,

  Does this solve bug 347 properly? (Do we now support the older versions
  of git?) I would just like to document this properly.
 
 Unfortunately no:
 
 $ ~/git/git/git --git-dir git/setup/.git --work-tree git/setup add HACKING
 Unknown option: --work-tree
 
 $ ~/git/git/git --version
 git version 1.5.2
 
 for 1.5.3 you still need to manually chdir() to the repo, perform the
 operation and then chdir() back. And that's already done by
 http://bugs.locamotion.org/attachment.cgi?id=125action=view ;-)

I am not sure, if I get the complete picture of the state of the git support.
Please correct me, if I am wrong about something:

1) the current implementation does not work without the --work-tree parameter
(e.g. with the two patches that were sent by Miklos (30.6.2008))

2) git v1.5.3 (or higher) behaves as expected - it is specified/documented to
need both --git-dir and --work-tree for any operation outside of the
current directory

3) for git v1.5.2 (and before) we have to change to the (base of the?) local git
repository before executing git


If these three assumptions are valid, then I would see the following
alternative solutions:

A)
- applying Miklos patches to the current git support
- adding a second module for git support (e.g. git-1.5.2.py) that would
include the ugly os.chdir workarounds. We have to keep both in sync until
we will finally remove support for git=1.5.2. The current module for the
Lucene indexing engine is splitted in a similar way, so I don't see problems in
implementing this.

B)
- adding the os.chdir workaround to the current git interface and removing all
occourences of --git-dir and --work-tree.

C)
- applying Miklos patches to the current git support
- ignoring incompatibility with git=1.5.2

Personally I would prefer (A) over (B) due to the cleaner code (in the
long-term). But of course, I would not mind (B), too.
The current version of git in debian stable seems to imply that (C) is not a
good option for us, or?

Any opinions?

Lars

-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-07-01 Thread F Wolff
On Wo, 2008-07-02 at 02:36 +0200, Lars Kruse wrote:
 Hi,
 
   Does this solve bug 347 properly? (Do we now support the older versions
   of git?) I would just like to document this properly.
  
  Unfortunately no:
  
  $ ~/git/git/git --git-dir git/setup/.git --work-tree git/setup add HACKING
  Unknown option: --work-tree
  
  $ ~/git/git/git --version
  git version 1.5.2
  
  for 1.5.3 you still need to manually chdir() to the repo, perform the
  operation and then chdir() back. And that's already done by
  http://bugs.locamotion.org/attachment.cgi?id=125action=view ;-)
 
 I am not sure, if I get the complete picture of the state of the git support.
 Please correct me, if I am wrong about something:
 
 1) the current implementation does not work without the --work-tree 
 parameter
 (e.g. with the two patches that were sent by Miklos (30.6.2008))
 
 2) git v1.5.3 (or higher) behaves as expected - it is specified/documented to
 need both --git-dir and --work-tree for any operation outside of the
 current directory
 
 3) for git v1.5.2 (and before) we have to change to the (base of the?) local 
 git
 repository before executing git
 
 
 If these three assumptions are valid, then I would see the following
 alternative solutions:
 
 A)
 - applying Miklos patches to the current git support
 - adding a second module for git support (e.g. git-1.5.2.py) that would
 include the ugly os.chdir workarounds. We have to keep both in sync until
 we will finally remove support for git=1.5.2. The current module for the
 Lucene indexing engine is splitted in a similar way, so I don't see problems 
 in
 implementing this.
 
 B)
 - adding the os.chdir workaround to the current git interface and removing all
 occourences of --git-dir and --work-tree.
 
 C)
 - applying Miklos patches to the current git support
 - ignoring incompatibility with git=1.5.2
 
 Personally I would prefer (A) over (B) due to the cleaner code (in the
 long-term). But of course, I would not mind (B), too.
 The current version of git in debian stable seems to imply that (C) is not a
 good option for us, or?
 
 Any opinions?
 
 Lars

Well, how far is the next Debian stable away? I would prefer A as well
if somebody is willing to do the work.

Friedel


-
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-06-30 Thread Christian Perrier
Quoting Miklos Vajna ([EMAIL PROTECTED]):
 On Sun, Jun 29, 2008 at 05:51:34PM +0200, Miklos Vajna [EMAIL PROTECTED] 
 wrote:
  I'll send a patch to correct this later tonight.
 
 Could you try this one, please?



As soon as possible, sure.



-
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-06-30 Thread Christian Perrier
Quoting Miklos Vajna ([EMAIL PROTECTED]):
 On Mon, Jun 30, 2008 at 09:08:27AM +0200, Christian Perrier [EMAIL 
 PROTECTED] wrote:
  The failure message changed..:-)
  [GIT] commit of ('/var/lib/pootle/di/level2/tasksel', 'debian/po/fr.po') 
  failed: 
 
 Heh. Are you sure there were actually changes to commit?
 
 I guess the problem is that commit failed because there were no changes
 to commit, and in this case the stderr is empty.

Yes, you're right, my mistake.

I was thinking I still had pending changes but I actually had zapped
the git cloned tree and recloned it. So nothing was left to commit

 
 If this is the case, I'm not exactly sure what would be the right
 solution, maybe check if stderr is empty and if yes, dump stdout instead
 stderr in the error message?


That would be better as, currently, trying to commit when nothing is
pending is indeed triggering erros like:

[GIT] commit of ('/var/lib/pootle/di/level2/tasksel', 'debian/po/fr.po') 
failed: 


-
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle


Re: [translate-pootle] Still problems committing with git

2008-06-29 Thread Miklos Vajna
On Sun, Jun 29, 2008 at 02:39:52PM +0200, Christian Perrier [EMAIL PROTECTED] 
wrote:
 Trace:
 
 2008-06-29 13:58:19: Traceback (most recent call last):
   File /usr/lib/python2.4/site-packages/Pootle/users.py, line 259, in handle
 page = self.getpage(pathwords, session, argdict)
   File /usr/lib/python2.4/site-packages/Pootle/pootle.py, line 523, in 
 getpage
 return indexpage.ProjectIndex(project, session, argdict, dirfilter)
   File /usr/lib/python2.4/site-packages/Pootle/indexpage.py, line 321, in 
 __init__
 self.handleactions()
   File /usr/lib/python2.4/site-packages/Pootle/indexpage.py, line 443, in 
 handleactions
 self.project.commitpofile(self.session, self.dirname, commitfile)
   File /usr/lib/python2.4/site-packages/Pootle/projects.py, line 599, in 
 commitpofile
 versioncontrol.commitfile(pathname, message=Commit from %s by user %s. 
 %s %
   File 
 /usr/lib/python2.4/site-packages/translate/storage/versioncontrol/__init__.py,
  line 292, in commitfile
 return get_versioned_object(filename).commit(message)
   File 
 /usr/lib/python2.4/site-packages/translate/storage/versioncontrol/git.py, 
 line 60, in commit
 raise IOError([GIT] add of ('%s', '%s') failed: %s \
 IOError: [GIT] add of ('/var/lib/pootle/di/level2/tasksel', 
 'debian/po/fr.po') failed: fatal: pathspec 'debian/po/fr.po' did not match 
 any files

Thanks.

The fix seem to be to use the --work-tree option as well:

Example:

$ git --git-dir git/test/.git add HACKING
fatal: pathspec 'HACKING' did not match any files

I think this is the current situation.

$ git --git-dir git/test/.git --work-tree git/test add HACKING

And this works fine.

I'll send a patch to correct this later tonight.


pgp9iV75zeEHZ.pgp
Description: PGP signature
-
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php___
Translate-pootle mailing list
Translate-pootle@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/translate-pootle